Kind of strange when someone who just joined this month knows what your degree is, but I think this would be quite hard. I wouldn't assume the weight and height of members here are normally distributed. The left tail of the distribution is probably very heavy, there seems to be a lot of very short members and very thin members so it would probably look like chi squared distribution, but it obvs wouldn't be because we aren't just a bunch of squared normal results. Having created a formula for this distribution, doing the necessary integration and creating estimators for the parameters we could create some confidence intervals for the mean height and weight. I'd rather create a multivariate linear regression to estimate an SF members weight based on their height and the size of items of clothes they've bought and sold and if they lift. We may have to use some time series for how long ago they bought the clothes. We could probably get an R value of like 0.8.
I think Superbobo has like a masters or Phd in stats so he'd probably be better at this.
you're drunk (or posing) if you think you'd get an R(squared) of .8 from any 'natural' data set.
you're drunk (or posing) if you think you'd get an R(squared) of .8 from any 'natural' data set.
The post is a joke, I don't think anyone here is creating a linear regression model on sf members height and weight. I just put a random number in, apologies for not using realistic numbers for my theoretical regression model. Please if you would like to go ahead and write in a more realistic number go ahead.
ah, fair enough...given the model you specified, weight as y, height and clothes size as x, think something in the range of .2 would even be optimistic. likely lots of error as well (ie, big confidence intervals around coefficient estimates).
I don't think fuji is wrong, actually. I think R^2 of .8 is quite reasonable, given how homogeneous the styleforum population is in terms of weight and height. I am assuming Mr Moo (an outlier way out in the right tail) is removed though.
fuji's first approach to create unbiased estimators is quite reasonable. I wouldn't a priori assume a distribution Chi^2 distribution. I would treat it as a gamma distribution, and do what fuji says.
lol, what? distribution of the dependent variable doesn't matter provided they're IID and the errors are assumed (and checked) to be normal.
Quote:
Originally Posted by ropseau2
Well, clothes size is a different issue altogether. We may safely assume 99.9999999% try to squeeze themselves into XS or S, no matter what their height and weight.
Take Reedobandito, for example. height 5'10", or thereaobuts. Weight 160lb +/- 20lb, if I read his posts correctly.
Size XS!!!!!
Then there's Mr Moo! Height 5'5", or therebouts. Weight 295 lb, by his own admission last Tuesday!
Size S!!!!!!
There will be absolutely no correlation between weight and clothing size!
I would be far more concerned about reporting error (a systematic bias in the sampling), because of Styleforum members consistently underreporting their weight and clothes size.
in order to get this thread back on track and flaming people for dressing poorly instead of whatever the hell is going on now, I am going to post an mc fit with "slippers" so you all can get back to doing what you do best
x-post
Quote:
Originally Posted by Ian Erickson
thrift fit from church today: andover shop, brooks brothers, henry rosen, j crew